AI handles 35 per cent of all NoBroker conversations: Founder Akhil Gupta

NoBroker founder Akhil Gupta says the firm's AI agent ConvoZen now handles 35 per cent of customer conversations, a number that may go up to 60-70 per cent in two quarters

Akhil Gupta, founder and chief tech and product officer of NoBroker, tells Avik Das about how the firm, which effectively cuts out traditional middlemen, is now adopting artificial intelligence as a way to converse with customers. The company's voice AI platform Convozen helps automate about 35 per cent of conversations with clients. The company is also targeting revenue of $10 million from this business as it onboards new customers. Edited excerpts:

 

How did ConvoZen came into existence?

 

In 2018-19, we started creating our own algorithms because most of the conversation was happening over calls. The idea was that if we can transcribe these conversations, which are multi lingual and with different dialects, we will be able to empower our process implementation adherence team (PIA) because we would be able to understand our customer far better. So then we started working on models where we can say 'this is our transcript, but give me the summary'. That also helped us understand if a customer was not happy with our service, or if my agent was mis-selling or over promising, and I could then give the feedback to them. That product was initially called CallZen which eventually became ConvoZen. The advantage is that I can now rate the call on different parameters.

 

What percentage of the conversations are handled by AI now?

 

It is about 35 per cent where the end-to-end conversation is handled by AI and all this has happened since the start of the year. And in some businesses, we have removed certain teams. So there was a team which used to check for rent-out properties, like whether it would be rented. Now we don't do that. Somebody was calling owners to get the photographs. Now the AI agent will automatically upload the pictures. In a voice AI mechanism, there are 3 parts which are understanding the voice, which is like speech to text, getting intelligence which is the LLM, and then speaking back to the customer in their own dialect. Because we have 10 years of data, we have trained our models.

 

How much do you think this can be scaled up to?

 

I feel genuinely 60-70 per cent of the conversation will be on AI and it will happen in the next two quarters. We are going a little slow because we want to ensure that the experience of our customer improves when we bring this particular technology.

 

When is this expected to become another business vertical for you?

 

We already have 50 customers which include HDFC Bank, HDFC Life, Maruti Suzuki, Spinny, and Cars24. Voice AI itself is a multi-billion dollar industry and we are doing close to about $2 million annually. The target is to have about $10 million by the end of March (2027).

 

With agents and humans working together, will there be any impact on headcount?

 

Every year, I would have added 15 per cent people, but with this technology, I may not add those. I don't think we will shrink too much but headcount may not increase drastically. With the same number of people, we will be able to generate more revenue. Some displacement will happen with the call centre people but the quality of work they do will also go up. This is because previously I may have attended 300-400 calls. Now with the agent, I may attend about 100 because the filtration is done by the agent and I know what exactly to speak so the quality of conversation goes up.